Is AI the future of Indie RPGs?

[Immortal]

Yet it cant even generate simple moderately quality/consistent assets for my game.

It can, you're just too lazy and stubborn to try to train a model and/or lora to do it.

My thesis is that it's worthless in the hands of people that suck. Either you just plain suck, or you're trying to use it outside a good use-case, which means you suck, or you don't integrate it into a larger process, which still means you suck. AI can produce assets for a pipeline just fine, and despite not investing much time into it I can generate consistent results and others with more tools can generate different angle renditions of the same character or item. Need textures for your project? It can do that. Need promotional art or still images? It can do that too. Enemies or wall art for your dungeon crawler, an art pass over your isometric pre-rendered backgrounds? Got it. What it won't do is to make you suck less.

In this case, he's exclusively talking about things like creating an interface or creating isometric tilesets.

His standard is also that it should be instantly useable without any work on his part to clean it up, or incorporate it into his design process.

In other words, he's not actually looking to use AI and is just trying to set a high standard that he feels is impossible so that he doesn't feel threatened as a programmer/amateur graphics designer.

Actual artists are already using AI in their creative processes. People, including myself (a non-artist), have used AI to create content for games, books, videos, internet ads etc.

But this guy is still in every AI thread on the site gaslighting himself that it's all a nothingburger. Sad.

I hit generate on a free website that runs an old stable diffusion and it didn't spit out my dream game in my head.

AI is shit and worthless.

 

[The_Sloth_Sleeps]

Yet it cant even generate simple moderately quality/consistent assets for my game.

It can, you're just too lazy and stubborn to try to train a model and/or lora to do it.

My thesis is that it's worthless in the hands of people that suck. Either you just plain suck, or you're trying to use it outside a good use-case, which means you suck, or you don't integrate it into a larger process, which still means you suck. AI can produce assets for a pipeline just fine, and despite not investing much time into it I can generate consistent results and others with more tools can generate different angle renditions of the same character or item. Need textures for your project? It can do that. Need promotional art or still images? It can do that too. Enemies or wall art for your dungeon crawler, an art pass over your isometric pre-rendered backgrounds? Got it. What it won't do is to make you suck less.

In this case, he's exclusively talking about things like creating an interface or creating isometric tilesets.

His standard is also that it should be instantly useable without any work on his part to clean it up, or incorporate it into his design process.

In other words, he's not actually looking to use AI and is just trying to set a high standard that he feels is impossible so that he doesn't feel threatened as a programmer/amateur graphics designer.

Actual artists are already using AI in their creative processes. People, including myself (a non-artist), have used AI to create content for games, books, videos, internet ads etc.

But this guy is still in every AI thread on the site gaslighting himself that it's all a nothingburger. Sad.

Why would I create a training set when it would be less effort to just make the assets? You don't seem to comprehend the inherent limitations.

I don't disagree that it is extensively used already to churn out crap. This is kind of the same reason I wouldn't use Unity3D. It's become associated with crap, and all the other things that go with it.

 

[Non-Edgy Gamer]

Why would I create a training set when it would be less effort to just make the assets? You don't seem to comprehend the inherent limitations.

Goal post shift.

First you say it "can't" do something. Now you say it's easier to do it yourself.

 

[The_Sloth_Sleeps]

Why would I create a training set when it would be less effort to just make the assets? You don't seem to comprehend the inherent limitations.

Goal post shift.

First you say it "can't" do something. Now you say it's easier to do it yourself.

Ok, even if it could do it right (it cannot), it would be less effort to make the images myself.

 

[Non-Edgy Gamer]

Ok, even if it could do it right (it cannot), it would be less effort to make the images myself.

Depends on how many images you need though, and whether you have the skill to make them already or not. If you needed hundreds, or if you had no idea how to draw isometric images, then something that can do the job for a low cost or free like this would be tempting. Especially once someone (besides the guy who already did it who others already linked to you) trains a model for it.

It's like you're looking at tractor and saying it's useless because you can do all your gardening by hand. Or saying that no one would ever hire an artist because you personally can already draw.

 

[The Wall]

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

 

[The_Sloth_Sleeps]

Ok, even if it could do it right (it cannot), it would be less effort to make the images myself.

Depends on how many images you need though, and whether you have the skill to make them already or not. If you needed hundreds, or if you had no idea how to draw isometric images, then something that can do the job for a low cost or free like this would be tempting. Especially once someone (besides the guy who already did it who others already linked to you) trains a model for it.

It's like you're looking at tractor and saying it's useless because you can do all your gardening by hand. Or saying that no one would ever hire an artist because you personally can already draw.

If you are going to create an analogy at least pick an accurate one. Take the robot vacuum cleaner, which surprisingly some people actually use. Much like AI art I think objectively these things are darned useless as you just have to clean up after them anyway, and they dont work for most houses.

 

[Immortal]

This is kind of the same reason I wouldn't use Unity3D. It's become associated with crap, and all the other things that go with it.

FYI - In the last year, Unreal is getting this reputation as well. Blueprint + Asset Store flipping is becoming extremely common in Unreal games on steam and a lot of people are bemoaning all the shit Unreal games flooding the internet (Many of them just unfinished ripped asset patreon-bait porn games lmfao).. so you better not use that engine either.

In fact.. you should write your own engine from scratch. Or better yet.. you should just quit.

If you are going to create an analogy at least pick an accurate one. Take the robot vacuum cleaner, which surprisingly some people actually use. Much like AI art I think objectively these things are darned useless as you just have to clean up after them anyway, and they dont work for most houses.

What are you trying to do again? An Isometric Tavern or something?

 

[The_Sloth_Sleeps]

This is kind of the same reason I wouldn't use Unity3D. It's become associated with crap, and all the other things that go with it.

FYI - In the last year, Unreal is getting this reputation as well. Blueprint + Asset Store flipping is becoming extremely common in Unreal games on steam and a lot of people are bemoaning all the shit Unreal games flooding the internet (Many of them just unfinished ripped asset patreon-bait porn games lmfao).. so you better not use that engine either.

In fact.. you should write your own engine from scratch. Or better yet.. you should just quit.

If you are going to create an analogy at least pick an accurate one. Take the robot vacuum cleaner, which surprisingly some people actually use. Much like AI art I think objectively these things are darned useless as you just have to clean up after them anyway, and they dont work for most houses.

What are you trying to do again? An Isometric Tavern or something?

I agree. I wouldn't touch that UE4 trash. TBH I haven't really kept up with commercial engines for the past 7-8 years but I not surprised UE4 has gone that way.
I am creating my own engine based on monogame.

 

[Immortal]

This is kind of the same reason I wouldn't use Unity3D. It's become associated with crap, and all the other things that go with it.

FYI - In the last year, Unreal is getting this reputation as well. Blueprint + Asset Store flipping is becoming extremely common in Unreal games on steam and a lot of people are bemoaning all the shit Unreal games flooding the internet (Many of them just unfinished ripped asset patreon-bait porn games lmfao).. so you better not use that engine either.

In fact.. you should write your own engine from scratch. Or better yet.. you should just quit.

If you are going to create an analogy at least pick an accurate one. Take the robot vacuum cleaner, which surprisingly some people actually use. Much like AI art I think objectively these things are darned useless as you just have to clean up after them anyway, and they dont work for most houses.

What are you trying to do again? An Isometric Tavern or something?

I agree. I wouldn't touch that UE4 trash. TBH I haven't really kept up with commercial engines for the past 7-8 years but I not surprised UE4 has gone that way.
I am creating my own engine based on monogame.

So you can't be fucked to collect 25 images online and use an autotagger to train a LORA for your needs in 2 hours.
But you will sit down and write a game engine from scratch?

You okay buddy? Does your handler know your posting on an 18+ RPG forum?

 

[Dexter]

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

This Finnish dude is putting together a Free 2.5D Point & Click Adventure using mostly AI Tools in the spirit of the challenge to see how far he can get and Blogging about it, although he has 20 years of experience as a 3D artist: https://echoesofsomewhere.com/

Here he talks about Backgrounds and Character modelling for which he used Midjourney (I believe there are more efficient ways of doing some of this by now). There's a few follow-ups for doing robots. Here is a post about Portrait generation. Here's a post about making scenes look less static and one about adding reflections. Here's one about how he animated the Mag train and AI-based smoke and fluid simulation. Here he's creating the logo.

 

[Egosphere]

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Yeah I think this is 100% correct. Arcanum 2 is probably within sight. There's still a lot of busywork that would be required, but imo, it's no longer intensive in technical skill. 2d art, textures, backgrounds, menus - all can be generated. Voice acting can also be generated. Even mundane writing just for background filler can as well.
Music remains the only hurdle, and i'm not sure for how long

 

[Trans-Financial-Man]

All I'm going to say is me and my team are aggressively pursuing a fork of Stable Diffusion with a custom sprite model for 2D and 2.5D art. We're a bit too far along to use it in our current project but using this may make my old "Daggerfall sequel but good" project a lot cheaper to make and far more inciting to investors because of that.

One thing I can see this sort of tool used for is getting an artist to make a handful of art in the style you want. And then training a model to make more. When you run into limitations you pay for another handful of art and train the AI on that. And that's before the innovations in this tech that will be coming in the next couple of years.

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Yes, this is the main reason I am excited for this tech. It will absolutely enable shovelware shit. So did Unreal and Unity when they made the tools so simple idiots could use it. But it will enable loads of people to make their dream game far cheaper and with far less reliance on other people. And that's a good thing.

 

[Immortal]

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Yeah I think this is 100% correct. Arcanum 2 is probably within sight. There's still a lot of busywork that would be required, but imo, it's no longer intensive in technical skill. 2d art, textures, backgrounds, menus - all can be generated. Voice acting can also be generated. Even mundane writing just for background filler can as well.
Music remains the only hurdle, and i'm not sure for how long

Using advanced AI text generation that can generate branching dialogue trees based on a few personality inputs (Alpaca / Pyg) you can actually make a proper 'talk to dead' spell too.

 

[The_Sloth_Sleeps]

All I'm going to say is me and my team are aggressively pursuing a fork of Stable Diffusion with a custom sprite model for 2D and 2.5D art. We're a bit too far along to use it in our current project but using this may make my old "Daggerfall sequel but good" project a lot cheaper to make and far more inciting to investors because of that.

One thing I can see this sort of tool used for is getting an artist to make a handful of art in the style you want. And then training a model to make more. When you run into limitations you pay for another handful of art and train the AI on that. And that's before the innovations in this tech that will be coming in the next couple of years.

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Yes, this is the main reason I am excited for this tech. It will absolutely enable shovelware shit. So did Unreal and Unity when they made the tools so simple idiots could use it. But it will enable loads of people to make their dream game far cheaper and with far less reliance on other people. And that's a good thing.

The Unity/UE4 comparison is an fair comparison and almost certainly how the future will play out.

We can look forward to massive amounts of shovelware for sure. Not something I'm excited about.

 

[Immortal]

All I'm going to say is me and my team are aggressively pursuing a fork of Stable Diffusion with a custom sprite model for 2D and 2.5D art. We're a bit too far along to use it in our current project but using this may make my old "Daggerfall sequel but good" project a lot cheaper to make and far more inciting to investors because of that.

One thing I can see this sort of tool used for is getting an artist to make a handful of art in the style you want. And then training a model to make more. When you run into limitations you pay for another handful of art and train the AI on that. And that's before the innovations in this tech that will be coming in the next couple of years.

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Yes, this is the main reason I am excited for this tech. It will absolutely enable shovelware shit. So did Unreal and Unity when they made the tools so simple idiots could use it. But it will enable loads of people to make their dream game far cheaper and with far less reliance on other people. And that's a good thing.

The Unity/UE4 comparison is an fair comparison and almost certainly how the future will play out.

We can look forward to massive amounts of shovelware for sure. Not something I'm excited about.

The internet is used by people to post stupid shit online.
We should probably not use it. Too much shovelware content.

 

[Aghoric Visions]

Its a nifty tool for those with no artistic talent

I keep reading this online but I don't think it's true and the two types of people that say this are either butthurt artists or clueless people who think artists are magicians. It's a tool that can be used by professional artists and those with more modest talents alike to supercharge their workflow, but those entirely without it will never produce anything all that great or even good with it. Most artists don't draw things freehand, or without reference. Comics were often just photobashing a page from an adult magazine or a lingerie ad together with a bunch of other photo and print material collected from whatever physical sources they had at hand which are then either copied or traced over. Concept artists already used the img2img workflow before AI was even a thing, they just did things manually, and I've seen art like this end up in the final game.



When Victor Pflug overshared about how the video game art sausage was made and people were disappointed the game art didn't spring directly from his creative genius but WW1 photos and stuff like that he responded with that actually making collages and twisting them around slightly is a high art form. No idea what his stance on AI is but it'd be hypocritical to be against it when it's just the same thing but slicker.

That whole art debate aside, knowing what is actually behind the artist kayfabe since camera obscuras were invented, you can see how untrue it is that nothing is required by the user just by looking at the latest post in this thread. Anyone can type in "Todd Howard as a cybernetic guy" into their model of choice but it's not going to be aesthetically pleasing. In fact, it might not look very cyborgy at all. But if you know precisely what you are after you can use AI in a tons of different ways in your process. You could photobash something together in photoshop, use img2img to let a model turn it more coherent, keep regenerating the parts you're not happy with. And it doesn't have to end there, you could trace over it, keep working on it in a digital painting program, turn it into pixel art Pflug style.

The reason so much AI art sucks isn't because the models are bad, it's because the people using them are. You could use it to generate a pulp cover for your fiction, drawing up a sketch for it to use and using the precisely right parameters, then refining the image in different ways until it is just right and fits what you had in mind writing the piece. Or you could just type in "super duper cool monster fighting giant monkey, artstation, funko pop, 4k, behance" and get the absolute shit your taste, knowledge and skill has earned you.

Meanwhile, I'm not that great at painting or drawing, but this is what even a little interest in different art forms gives you despite Microsoft's version lacking regeneration tools and img2img and me investing the bare minimum of time and effort...

What a passing interest in film gives you.

Architecture and photography

Comics

General illustration

Cybernetic Todd Howard didn't need to look like crap and I don't think the tools can be blamed for it. Sorry Zed, I'm not trying to be mean, I'm just illustrating my point. For some people it ends with the kind of stuff I posted, and it might be good enough for their purposes, but it doesn't need to end there and the guys and girls that can improve results like these are the ones that can benefit the most from them.

Yet it cant even generate simple moderately quality/consistent assets for my game. I grow very tired of the whole "you havent got the prompt right" or "anytime soon" arguments that the proompter enthusiasts keep regurgitating over and over. Its capacity to me appear to be incredibly limited, that is to say, I have no doubt it will be widely employed to make rubbish - but beyond that the uses are limited.

I can't draw so much as a stick figure, so yes. Artists are magicians to me. All in all I think the tool is a good thing. If it helps level the playing field for the uninitiated that's great.

 

[Not.AI]

I firmly believe that couple of devs, who are ok-good programmers and creative souls without technical knowledge of art making, can with current AI tools make their own little dream adventure-RPGs by the end of this year, already. Art? 3d/2d art? Animations? Voices? Coding Co-Pilot? Check. Check. Check. And check

WHAT THE FUCK IS STOPPING US FROM MAKING NEW ARCANUM or FALLOUT 3?! Nothing except human laziness

Seconded. 100 percent this is the case - human laziness is as usual the main culprit.

A couple people can even make competitors to the so-called best-in-class ai tools themselves, without too much expense, see for example the recent Open Assistant.

 

[Dexter]

https://blog.luden.io/generated-adv...-and-midjourney-prompts-included-f87e7e615204

 

[Non-Edgy Gamer]

Impressive, but still using a lot of hacky attempts to adapt out-of-the-box Stable Diffusion generations into game design rather than do anything innovative. It was interesting how they generated music based on PoE2 though.

There's an advantage to this over outsourcing in price and speed, but the overall product is either worse or isn't any better than someone doing it by hand. E.g., there's not going to be true consistency, since he's generating entire areas instead of generating a tileset like he should be doing.

I think we need to move beyond proof of concept at this point. I want generative models that can create thousands of sprites for a roguelike and pair them with GPT generated descriptions. I want examples of AI art that are either indistinguishable from human art, or plainly better.

 

[Non-Edgy Gamer]

anvi, I don't think it's optimism. It's possible using a custom model and GPT-4. It's simply that no one has trained the model or setup the GPT-4 integration. It's possible without the model (less consistent results) or the integration too. I can do it by hand. But the automation would be nice.

Pipeline is: GPT-4 > game formatted description > AI formatted prompt > AI generated image > ML to make the BG transparent

 

[The_Sloth_Sleeps]

https://blog.luden.io/generated-adv...-and-midjourney-prompts-included-f87e7e615204

Zoom in and marvel at how turd it looks.

 

[Non-Edgy Gamer]

There's really no excuse for it in a paid product when you can fix stuff with inpainting or photoshop/krita SD integrations.

 

[KeighnMcDeath]

Can you interact with EVERY object?

 

[Non-Edgy Gamer]

Can you interact with EVERY object?

No, but it'd be fairly easy to make it so. It's just an image. Edit it to make whatever part interactive. Select desired piece, copy, use AI to fill in the background where the part was. They explain in the article they do something similar for passing behind objects.

But as I was saying, this is inefficient and unprofessional. Better to render real tileset elements and not be lazy.

edit: oh, and having a lot of object clutter can be confusing for a game. You wouldn't want them all to be interactive. In fact, just the fact that so many useless buildings and objects exist in these images is a negative.